Vax Replication - Local or WAN

[Phillip Bruce_2]

Has anyone succesfully performed Storage Replication both locally and Over the WAN?

I know how to setup shadow which is nowdays called mirroring. But say I have 2 separate clusters and I want to replicate the data to the other cluster so the databases are in sync. Even want to do this over a WAN if possible or even
maybe a direct SAN implementation.

Also has anyone Tied a SAN to the Vax CI with the Star Coupler? Can it be done?

[Alex Daniels]

It is possible to do this, but it is highly dependant on what kind of performance you are seeking and the latency of your network configuration.

There are a number of ways to do this, here are some, but it is so dependant on your requirement..

1.
Simplest way would be to merge the two clusters and MSCP server the disks and/or have a local copy shadowed.

2.
Your could also look at using DFS and keep your clusters seperate, but you would need to purchase the software and the setup is not trivual.

3.
NFS would be my least favoured option, as it almost certainly wouldnt have the performace and you may have issues with locking.

4.
You could even just access the disks remotly using the full path including node name from the remote cluster.

Having remote shadow members within a cluster across a WAN, however is pretty stright forward. I would always look to have seperate non-wan shadowed system disks for performance, but otherwise it should work well.

I think though you need to look more into what kind of performance you are looking to get and what your requirements are. As the best option really depends on this and your budget.

100mb/sec cards are now available for VAXen and this may be a cheap option, for more performance.

[Martin Johnson]

If you are using a database product like Oracle or Sybase, they have replication services available (for a price) that will replicate locally or over a WAN.

HTH
Marty

[Keith Parris]

There are quite a few ways of doing data replication, both locally and over a WAN. Both software- and hardware-based solutions are available.

In hardware, controllers like the EVA, HSG80, or XP series take data on a local disk and replicate it to a disk attached to another similar controller at another site. VMS nodes can access the data only from one site at a time; you tell the controllers, typically via a CLI, which site to present the data at, and which site will contain the replicated copy; if you need to fail over, CLI commands are typically used in performing that failover.

In software, data replication can be done at the operating system level, as with Host-Based Volume Shadowing software; in the database software itself, or by taking snapshots of the database and copying logfiles across between systems in a method known as "log shipping"; or by replicating transactions in middleware with something like Reliable Transaction Router (RTR), which comes free with VMS. You didn't say what database software you use, but for example Oracle Server has DataGuard (earlier called Hot Standby) available, and Oracle Rdb also has a similar capability.

To use Host-Based Volume Shadowing, either the systems at both sites would need to be in the same cluster, or you'd use a cross-site SAN (either over dark fiber or by using boxes from companies like CNT, SAN Valley, or SAN Castle to bridge the Fibre Channel over an ATM link, LAN connection, or IP network) and use Selective Storage Presentation to present a given disk to only one of the two clusters at a time (so the replicated copy is only accessible to one cluster at a time to avoid data corruption, but can be presented to the opposite cluster when needed for failover). If data replication is done at the database or middleware layers, the clusters can remain separate.

There's more info on data replication in my DECUS presentation "Disaster Tolerant Cluster Technology and Implementation" found at http://www2.openvms.org/kparris/

What sort of distance is there between the sites?